Engineering Open Problems: Choose an Interface to Improve
Quantum engineering research asks how imperfect components can deliver a reliable capability together. The open problem is often an interface: a decoder misses a deadline, a transducer heats a processor, or calibration interrupts a workload. A component record matters when it improves a complete task under realistic operating conditions.
Helpful background. Resource estimation connects operation requirements to system cost.
Route by the interface that fails
Section titled “Route by the interface that fails”| Interface problem | Topic owner | Measurable contribution |
|---|---|---|
| Syndrome data arrive faster than decisions | Real-time decoding | Accuracy with bounded backlog and latency tails |
| Useful qubits disappear under simultaneous operation | Hardware scaling | Yield- and drift-aware logical capacity |
| Logical operations exhaust routing or factories | Fault-tolerant architectures | Lower complete workload cost at fixed failure probability |
| A processor cannot use an optical link | Transduction | Low-noise, end-to-end quantum channel under operating load |
| Remote modules wait too long | Modular computing | Application runtime including entanglement retries and memory noise |
| Network nodes disagree about a resource | Network protocols | Verified semantics under resets, delays and failure |
Established: integrated demonstrations and efficient subsystems exist in specified regimes. Active: sustaining their performance when operated together and at larger scale. Conjectural: an architecture forecast assembled from independently optimal component records. Sparse Blossom and the distributed-ion experiment illustrate respectively algorithmic and hardware evidence for interfaces. Decoder paper, remote-computing experiment.
Write a contract for each connecting arrow
Section titled “Write a contract for each connecting arrow”Specify rate, latency distribution, error channel, memory, power or thermal load, availability and recovery behavior. A mean value is inadequate when rare faults dominate a long workload. State which parameters were measured simultaneously and which are projections.
For a stream with arrival rate and service rate , is a basic stability requirement in an appropriate stationary queue model. It says little about a hard deadline unless the service-time tail and burstiness are known. Similar distinctions recur in photon supply, calibration and resource factories.
Entry project: a bottleneck sensitivity model
Section titled “Entry project: a bottleneck sensitivity model”Select a small logical circuit or entanglement service. Build a discrete-event model from one compatible set of published component parameters. Vary one interface at a time and identify the improvement that most reduces delivered cost. A laptop suffices; hardware access is needed only when validating the model’s uncertain interface.
Deliver a reproducible schedule, parameter table, runtime distribution and failure budget. Compare with a serial baseline and a physically meaningful lower bound. Failure tests include outages, doubled feedback delay, reduced yield and correlated errors. Stop optimizing a subsystem once another becomes dominant. Research begins when a new design or control policy improves the full task under measured conditions, or when the model identifies a previously omitted requirement verified by experiment.
Novelty and resource checks
Section titled “Novelty and resource checks”Confirm access to the controller, fabrication, optics or calibration traces needed for the proposed test. A project requiring a future device can still be valuable theory, but its deliverable must be a conditional bound or design, not an unperformed demonstration. Reviewed 11 October 2026; selected evidence is from 2025 publications. Topic owners hold the more detailed assessments.